Object-Based Probabilistic Forecasting of Severe Weather Hazards in the 2018 HWT Spring Experiment: A Multi-Variable Probabilistic Approach for Research and Operations.

Verification of high amplitude, small-scale features such as precipitation systems in convection-permitting models has been shown to benefit from object-based approaches. The value of object-based verification lies in the ability to quantify storm attributes of greatest subjective interest to forecasters despite small spatial displacements. However, a common concern about the object-based approach is that it relies on many subjective user-defined parameters. Furthermore, previous object-based verification studies have typically focused on deterministic forecasts and/or a single variable such as precipitation, reflectivity or updraft helicity to define objects and object attributes. The advantages of an object-based approach may be even more pronounced in the context of ensemble probabilistic forecasts involving multiple variables that are relevant for severe weather forecasting, given the large data volume involved.

This study simplifies the object-matching process by limiting the user-defined parameters to a few physically meaningful values. The object-based verification is also applied in a new ensemble probability framework that accounts for multiple variables and severe weather hazards. Ensemble forecasts from the 2017 HWT Spring Experiment will be used to demonstrate agreement between this objective approach and detailed subjective ensemble evaluations. Results from an ongoing systematic verification study aimed at better understanding optimal ensemble design will then be presented.

An object-based probabilistic forecast interface to the OU MAP (Multiscale data Assimilation and Predictability) lab’s contribution to the CLUE (Community Leveraged Unified Ensemble) is being provided to forecasters during the 2018 HWT Spring Experiment. Results and forecaster feedback relating to the use of such an approach to convection-permitting ensemble post-processing in the context of real-time severe weather forecasting will also be presented.